> discussion is a waste of bits.
rob: with respect, that is your opinion. and, with respect: i am
compelled from experience to point out that your opinion tells me that
you simply do not have any experience in the scale or scope of the
task that you are facing.
with respect: you are failing in your duty of care and responsibility
to the netsurf community if you do not fully comprehend the various
options out there, and have "dismissed them" for any undocumented and
improperly-evaluated reason.
there are damn good reasons why Common Object Models are deployed for
the purpose described, and you have not even _begun_ to describe any
technical reasons - of any kind! - as to why you are dismissing the
options being presented.
should you wish to develop netsurf entirely "in secret" behind closed
doors please feel free to do so but should you choose to do so please
actually state on the web site "external technical input from
experienced software developers is not welcome in this project".
i trust that this is not the kind of image that the netsurf project
wishes to present to the outside world.
> NetSurf will almost certainly never rely
> on a heavyweight tricky-to-port fat library such as any of the Glib
> family, or anything else like that. If we need something as you
> suggest, it will probably be our own, and not just because of NIH
> syndrome, but because we have special requirements.
1) where may a list of the special requirements be found, such that i
may review them and thus focus spending my personal free time and
personal funds aiding and assisting the users of netsurf more
productively?
2) with respect: why did you not raise this earlier rather than
letting us spend large amounts of time discuss matters which in your
opinion are "a waste of time"?
3) to dismiss existing Common Object Models in general as
"heavyweight" is pure foolishness, rob. and is disrespectful towards
people who wish to aid and assist the netsurf community.
3) have you done a full technical evaluation of the amount of code
that is likely to be generated by "rolling your own" Common Object
Model or Code-Generator?
4) the discussion is not limited to gobject. as part of the
discussion, a number of alternative COM-inspired technologies were
found, many of which were designed with embedded systems in mind.
have you evaluated those technologies?
i actually *know* what you're facing, rob, to achieve the goal. it is
simply - without fail an absolute undeniable 100% cast iron guaranteed
inescapable fact that you are either a) *GOING* to have to use or
write a Common Object Model system, or a Code-Generator.
* a code-generator *will* add hundreds of thousands of lines of code
to the netsurf project. this is an inescapable undeniable fact.
* writing your own Common Object Model *will* take you something like
an entire man-year to write and get right.
the question is: is it _really_ worth it, and, once the task has been
completed, are you *absolutely* sure that it will result in
"fulfilling the special requirements" (*1), by virtue of you having
done a full and comprehensive analysis of the execution speed, memory
usage and binary object size?
(*1 which you didn't actually list. or take the time to point anyone
towards a document which describes them.)
i have absolutely no doubt that should the netsurf team choose the
route of "rolling their own", it will be one of the most superb free
software COM systems in existence (i know the professional and
technical experience of the netsurf developers is exceptionally high).
should the netsurf team choose that route, i look forward to being
able to take that excellent software (libre) licensed code so
developed in order to utilise it in other projects, and i most
certainly will evaluate it and advocate it right across the board in
other forums which could benefit from the work done.
l.
No comments:
Post a Comment